In France, while more than 20% of dwellings belong to the private rental sector, the understanding of this market is still imperfect and is mostly limited to specific areas (e.g. Paris). The creation of the Observatoires Locaux des Loyers (OLL) network from 2013 intend to address this problem by providing to researchers high-quality detailed microdata of dwellings structural characteristics for several areas in France (Lille, Marseille, Toulouse, etc.).

In collaboration with the Ministère du Logement et de l'Habitat Durable (French housing ministry) and the Observatoire des Loyers de l'Agglomération Parisienne (OLAP) which manages the OLL’s, we are the first to capitalize on the OLL’s and OLAP work, i.e. to use their databases from 2014 to 2015 to improve the understanding of the private rental sector of all areas where there is an OLL. Specifically, we present in this paper an extensive analysis of the factors that determine the rent levels in that sector in France, based on microdata of dwellings structural characteristics provided by 14 OLL and the OLAP.

For this purpose, we use a hedonic regression model in which dwelling rent is characterized by a bundle of several characteristics. In addition to the dwellings’ physical characteristics (type of dwellings, number of rooms, total surface etc.), we use a set of variables that includes neighborhood characteristics (median income, accessibility, etc.) and a set of variables that relates to environmental quality (air pollution and proportion of green areas or forests). Since it is widely known that dwelling rents tend to be heteroskedastic and spatially autocorrelated, and that in the presence of such a phenomenon, the ordinary-least square estimator of the parameter of the hedonic model is inefficient, we use the nonparametric heteroscedasticity and autocorrelation consistent OLS estimator (OLS-SHAC). This estimator is robust against possible misspecification of the disturbances and allow for unknown forms of heteroscedasticity and spatial auto-correlation between dwelling rents.

In addition to the exploration of the determinants of dwelling rents, we propose a typology of the French private rental sector: using principal component analysis and hierarchical clustering, we find clusters of areas for which rent levels are explained by similar attributes (in term of magnitude). Our typology may be used to design areas-based policies rather than one for the entire French market.